Ligand-activated Cre recombinases are widely used for studying gene function in vitro and in conditional mouse models. To compare ligand-dependent Cre recombinases, different Cre estrogen receptor fusions were introduced into the ROSA26 locus of embryonic stem (ES) cells and assayed for genotoxicity and recombination efficiency. Of the tested recombinases, the CreERT2 variant showed no toxicity and was highly responsive to ligand induction. To constitutively express CreERT2 in mice and also to clarify whether the CreERT2 system displays background activity, we generated a knock-in mouse line harboring the CreERT2 coding region under the control of the ROSA26 locus. Analysis of this ROSA26-CreERT2 deleter mouse with different reporter strains revealed ubiquitous recombination in the embryo and partial recombination in peripheral and hematopoietic tissues but no effective CreERT2 expression in the brain. Furthermore, using flow cytometry, we found low-level background recombination in noninduced bitransgenic ROSA26-CreERT2/EGFP reporter mice. To determine whether background activity poses a general problem for conducting conditional in vivo experiments with the ROSA26-CreERT2 deleter, we used a sensitive conditional skin cancer model. In this assay, cancer induction was completely restricted to induced bitransgenic CreERT2/K-Ras(V12) mice, whereas noninduced control animals did not show any sign of cancer, indicating the usefulness of the ROSA-CreERT2 system for regulating conditional gene expression in vivo. The ROSA26-CreERT2 deleter strain will be a convenient experimental tool for studying gene function under circumstances requiring partial induction of recombination in peripheral tissues and will be useful for uncovering previously unknown or unsuspected phenotypes.